Synthesis of separation sequences using a genetic algorithm

This paper deals with the implementation of a non elitist genetic algorithm with roulette wheel selection for the synthesis of separation sequences. After a brief recall of problem combinatorics, the coding strategy based on the notion of stack for separating a list of elements with feasibility constraints, as well as the genetic operators used, are detailed. A didactic example is first presented for estimating the numerical values of the algorithm basic parameters, and then two large scale examples concerning the separation of mixtures involving 14 and 16 hydrocarbons by means of distillation sequences, are reported. The obtained results, compared with those given by a simulated annealing procedure and a purely random search, clearly show the genetic algorithm efficiency.